rscole@evergreen.edu
Over the past eight years I have convinced myself that it is entirely possible, indeed, highly desirable, to teach mathematical modeling to students at the college algebra level, and in every math course subsequent to that. This involves concepts of differential equations, but can be done without mentioning calculus. What makes it possible is mathematical modeling software like Stella . This software enables students to focus their attention on how parts of a system interact (by building visual diagrams), while the software builds the coupled difference equations associated with the flow diagrams. Stella allows the teaching of a level of mathematical sophistication far beyond that associated with college algebra or pre-calculus. Indeed, most students emerge from my learning communities with the basic ideas of systems dynamics and systems thinking.
Here’s a brief outline of what's involved. The fundamental ideas concern amounts (something that can accumulate in a given location), and a flow between the various points of accumulation.
The significant thing about this characterization is that students bring an intuitive grasp of inflows, outflows and amounts with them -- they have direct experience with water flowing into and out of a kitchen sink, or a bathtub. We capitalize on this intuitive grasp to develop a more sophisticated understanding of how dynamic systems change in time.
Most students have a loose grasp of the fact that the future amount of anything in a bucket equals the present amount plus the inflows minus the outflows:
Future Amounts = Previous Amounts + (Sum of Inflows - Sum of Outflows)*Time
The significant thing about this characterization is that students bring an intuitive grasp of inflows, outflows and amounts with them -- they have direct experience with water flowing into and out of a kitchen sink, or a bathtub. We can capitalize on this intuitive grasp to begin to develop a more sophisticated understanding of how dynamic systems change in time. The crucial thing is that students at the college algebra level can begin to mathematically probe the time evolution of systems. Normally this is typically deferred until integral calculus, and isn't studied in depth until the differential equations. If we use any of a myriad of environmental topics as the applications, we end up teaching a significant amount mathematical modeling of environmental systems, and I've found that college algebra-level students can grasp a lot regarding the behavior of systems of differential equations. Most students get excited because they see the applications of mathematics to larger issues that they care about.
Jazzing up the notation a bit, we can write:
A
 t
  t
   
Sum of Inflows

Sum of Outflows

  t
Or, converting to derivative notation: dA dt
 lim

 t

0
  
Inflows
 
Outflows


The above equation is really the basis of all that follows. To be sure, we may end up rather quickly examining a system of coupled ordinary differential equations (ODE’s), but the above expression is the fundamental idea for a significant amount of modeling.
In Stella flows of things (Amount per unit time) flow are called rates and are represented by the symbol:
Amounts are called stocks and are represented by the symbol:
Stock
Stella also uses converters to input parameters, or do arithmetic operations, unit conversion, or other mathematical necessities. Connectors are the Stella devices that carry information flow between these components. They can be thought of as wires that carry information. This is in contrast to the double-lined flow arrow, which is best thought of as a pipe that carries stuff from one location to another. conv erter f low
These four components, flows, stocks, converters, and connectors comprise the fundamental building blocks for systems modeling. All models can be constructed from them.
Five Elementary Behavior Patterns of Dynamical Systems
Five elementary patterns of system behavior emerge, and are descriptive of a wide array of environmental phenomena. They are: (1) linear growth or decay, (2) exponential growth or decay,
(3) logistic growth, (4) overshoot and collapse, and (5) oscillation. Illustrations follow.
Linear Growth or Decay An example might be the individual consumption of food on a long cruise aboard ship (where the food supply is fixed at some initial value, and additional food from outside sources is not available). Assuming an individual eats a constant amount of food each day, we could model the amount of food in the ship's stores as: where we have taken the constant rate of consumption, k , to be positive, and explicitly included the minus sign to indicate that
A ( t ) decreases with time. dA
 dt
 k
The associated Stella diagram would be: where the consumption flow would be equal to k .
Ship's Store of Food consumption
We all know the solution to this differential equation is: A ( t ) = A (0) - k * t
Stella would plot a graph of this solution.
Exponential Growth or Decay Population is the easy example here. Exponential growth is characterized by the rate of change being proportional to the amount. Hence, where r is the net growth rate (birth rate minus death rate). dN dt
 r

N
The associated Stella diagram would be:
The solution to this ODE is N ( t ) = N (0) * e rt and again Stella would plot the solution. birth rate births
Population N deaths death rate
2

This model is the visually identical to one of super- or sub-exponential growth: where q is typically a small quantity. dN dt
 r

N
1
 q
Stella distinguishes the exponential from super-exponential growth by the form of the equation typed into the flow symbols. However, since super-exponential growth leads to an infinite amount in a finite amount of time (contrast with plain, garden-variety exponential growth!) this function is worth detailed investigation. Worldwide consumption of petroleum from 1900 to 1970 was best fit by a super-exponential curve. So was world population from 1900 to 1990!
Logistic Growth No study of exponential population growth is complete without including the concept of carrying capacity. This introduces a non-linearity into the above linear ODE. The most common way to introduce carrying capacity is as follows: dN dt
 r

N


1
N
K where K is called the carrying capacity .
The associated Stella diagram would be:
The solution, which can be obtained in births
Population N deaths closed analytic form, is the characteristic
S-shaped curve studied in population birth rate death rate carry ing capacity K biology and elsewhere.
We've pretty much exhausted the typology of solutions to single first-order differential equations.
However the number of applications is tremendous; we could spend half a year on these topics alone!
Overshoot and Collapse When a population is partially or totally dependent on a nonrenewable resource, the potential exists for a catastrophic collapse of the population as its consumption of the finite resource increases. This could be modeled by the following system: dN dt
 b

N


 1

R ( t )
R ( 0 )



N where b is the per capita birth rate, resource at time t , and R
R ( t ) the amount of the
(0) the initial amount of resource. dR
 c

N ( t ) dt
The associated Stella diagram would be: where c births
is the per capita consumption rate, and population at time t .
Population N deaths
N ( t ) is the death rate birth rate
There are multiple variations of this model, but the all lead to a solution where the population
Resource grows rapidly at first, peaks, and falls quickly to very close to zero. consumption rate consumption
For some reason, students get very interested in the behavior of this model, and want to try out many of the variations of the functional form and the model parameters. They'll be the leaders in the twenty-first century!
3

Oscillatory Solutions Certainly studying the simple harmonic oscillator is a worthy endeavor in ODE classes, and in a dynamical systems class. However, I'll illustrate a different set of two coupled first-order equations, ones that describe a predator-prey relationship. Ecologists start with the classical Lotka-Volterra equations: dV dt
 r

V
  
V

P where r is the inherent net growth rate, and
α
the coefficient of predation for the victim species. dP dt
  
V

P
 d

P where d is the inherent death rate for the predator, and β is coefficient of predation for the predator species.
The associated Stella diagram would be:
The solution to this set of equations gives the
Victim V pop growth predation classical Lotka-Volterra oscillation of each species. The phase-plane diagram of this model is a closed figure, but adding a inherent growth rate
Predator P alpha succession of more biologically realistic terms to the model results in a mathematical typology of phase-plane attractors. growth f rom eating deaths beta inherent death rate
Incidentally, a fascinating account of the efforts surrounding Vito Volterra’s efforts to understand Mediterranean fisheries after WWI (which resulted in his version of the Lotka-Volterra equations) is contained in Sharon E. Kingsland’s book
Modeling Nature (U. of Chicago Press, 1995, 2 nd
edition).
These five basic behavior patterns comprise the types of responses that more complex environmental models can yield. I typically examine perturbation from equilibrium states, and probe more deeply the types of oscillations that can occur, particularly exploring types of attractors students can expect to see in the phase plane plots.
Environmental Applications I have used these modeling techniques with students at the college algebra level and in upper division science learning communities. What changes is the mathematical sophistication of the students, and the level expectations I have of them. I've developed computer laboratory exercises exploring pollution concentrations as functions of time in simple lake systems, and some models of age-structured populations of salmon in Pacific northwest rivers. In addition I've developed lab projects investigating population biology, including the concepts of exponential and logistic growth, and predator-prey relationships. I developed a lab illustrating bioconcentration of lead in human bones, and another investigating the spread of epidemics in human populations. I’ve also developed simulations involving biogeochemical cycles, trophic-level food webs, and groundwater flows. The range of potential environmental, ecological, health-related, geological, or biological applications is almost endless.
So What ?
So what's the big deal about teaching all this stuff to college algebra-level students?
Why bother? First, any potential scientist, regardless of field, should learn the techniques of modeling dynamical systems. This will be an essential scientific skill in the twenty-first century.
4

Second, for any student who does not wish to go into any of the scientific fields, systems thinking becomes an important skill of citizenship in the twenty-first century.
But aside from these high-sounding career-related reasons, why should a mathematics instructor care about teaching dynamical systems as a part of college algebra? The primary reason is that the level of conceptual and mathematical thinking that emerges from a study of dynamic systems is far more profound than the level of thinking that emerges from studying algebra. When you step back from it, college algebra is really a dreary subject. We teach things like order of operations, rearrangement of expressions, graphical representation, solving equations, typology of equations
(linear, power, exponential, etc.
) and, in the 1990's, fitting curves to a set of data points. Don't get me wrong, these mechanical skills are essential. None of us could teach dynamical systems without most of them. But the intellectual content of algebra is mostly about syntax and rules of grammar .
Most folks don't get too excited about those topics when studying English, let alone when studying the foreign language of mathematics.
By contrast, dynamical systems forces us to think about dynamic processes. Algebra is static.
Dynamical systems, by definition, are not. Process is one of the things that calculus deals with, and so dynamical systems become excellent preparation for calculus. Dynamical systems force us to look at the big picture. Before we build a model, we have to examine a global description of the processes at work and then begin to move toward the specific. Where in algebra do we do that?
Systems thinking really forces us to think of processes as in closed loops (Spaceship Earth metaphor!), even if we don't model the entire loop. Dynamical systems introduce the concept of flow of information and of feedback loops. Where do we ask our students to experiment with feedback loops in algebra? Dynamical systems software gives immediate feedback about students mathematical model building. It is easy to get runaway processes in models, but very few such processes exist in nature, or in what we call the "real world." This forces students to pay attention, when modeling, to the checks and balances that prevent runaway processes. Moreover, dynamical systems often describe something in students' lives that they really care about. Engagement becomes automatic. Contrast this with trying to convince a student that syntax and rules of grammar actually matter to him or her. Finally dynamical systems force us to look for causal factors -- systems don't "just behave that way, without a good reason." This encourages students to start making connections between things, often very disparate things. Systems thinking requires students to think precisely and creatively beyond syntax and grammar, even though they have to use this syntax and grammar to make meaningful models!
What kinds of intellectual concepts do (most of) my students gain from studying environmental systems? Most understand that a closed system will reach equilibrium if left alone for a while, and that this equilibrium state is independent of the initial values of any of the stocks. Most understand how to test for the stability of such an equilibrium. They also understand that if you add a finite pulse of one stock ( e.g.
pollution) to a closed system, a new equilibrium for the system will ultimately be reached. Most of them have a clear picture of how flows work in compartment models, and what characteristics of those flows lead to a massive build-up of a substance in a given stock (they understand bio-concentration in the food chain!). Virtually all of them recognize the relationship between concentration of a pollutant and the flushing time in a given reservoir. They also understand the idea and meaning of the residence time of a substance in a stock when material is flowing through that stock. Most students see what causes overshoot and collapse in a system, and what causes a system to oscillate. Some few of them even understand what characteristics lead a system to exhibit chaotic behavior. Most of them have a reasonably good idea of what a
5

mathematical model is, and what its limitations are, and almost all of them have become fairly adept at finding the assumptions of a given model.
What don't they know much about, as a result of working with me? They don't know much, if anything, about conic sections, factoring, or the Law of Cosines, nor do they get a lot of practice reducing arcane algebraic expressions to so-called standard form.
But then, you know that my biases will be with teaching students skills that they will need to understand the environmental and ecological issues they will face in the twenty-first century!
Good Books on Environmental Modeling That Use Stella
Modeling the Environment , by Andrew Ford, (Island Press, 1999), ISBN 1-55936-601-7
‘Probably the best text available for introductory modeling, the book gives an excellent overview of all aspects of modeling. I’ve used it multiple times in modeling classes. It has several excellent case studies and a fine web site for students and faculty alike at www.islandpress.org/ford
Dynamic Modeling of Environmental Systems , Michael L. Deaton & James J. Winebrake, (Springer-Verlag, 1999),
ISBN 0-387-98880-7
The book is excellent. It contains more mathematics than Andy Ford’s book, being aimed at students who have had one semester of calculus. However, the book is quite accessible to students, even if they haven’t had calculus.
It has an accompanying CD-ROM, which contains a lot of models for students to work with. I have used this book in my Freshwater Ecology program.
Consider a Spherical Cow , John Harte, (University Science Books, 1988), ISBN 0-935702-58-X
This is probably the original environmental modeling book, published back in the era of pencil and paper mathematics. It is an excellent source of potential projects and homework exercises. Besides, its appendices contain the BEST source of environmental data to use in problems that I’ve ever seen. Where else might you find the volume of the Indian Ocean, the mass of the stratosphere, or the sulfur uptake of plants on the continents?
The Dynamic Environment , Leonard Soltzberg, (University Science Books, 1995), ISBN 935702-37-7
This book has the Stella models to accompany Consider a Spherical Cow . It is a good reference, but is not a textbook.
Modeling Dynamic Biological Systems , Bruce Hannon & Matthias Ruth, (Springer-Verlag, 1997), ISBN 0-387-94850-3
This is a great collection of Stella models on a very wide variety of topics, but it feels like a collection of mathematical models around which Stella was wrapped as an afterthought. ‘Not suitable as a text, but a great resource of ideas. It has a companion volume called Modeling Dynamic Economic Systems ,
6